Hormone-dependent Tumor Regression in Vivo by an Inducible Transcriptional Repressor Directed at the PAX3-FKHR Oncogene¹

Chromosomal translocations that result in the creation of chimeric transcription factors that deregulate specific target genes are a type of genetic alteration frequently associated with oncogenesis(1). Different modes of deregulation include gain or loss of transcriptional activation or repression function. The model system,ARMS³, chosen in this study provides a clear example for this phenomenon for human pediatric solid tumors (2).

ARMSs occur due to a highly specific chromosomal translocation event[t(2;13) (q35;q14)] that juxtaposes the DNA-binding domains of PAX3 with the transcriptional activation domain of FKHR (3). In transfection assays, PAX3-FKHR, the causative oncogene in ARMS, functions as a more potent activator of transcription than the wild-type PAX3 (4). Less frequently, another translocation[t(1;13) (p36;q14)] fuses the PAX7 DNA-binding domain to the same FKHR activation domain (5). These studies suggest that these PAX proteins have sustained a gain of function that leads to ARMS tumorigenesis (2). However, the PAX3-FKHR-activated target genes responsible for ARMS have not been defined.

The PAX gene family consists of nine members that are unified by the presence of the paired box DNA-binding domain and are subclassified based on their genomic organization. PAX proteins play regulatory roles in pattern formation during organogenesis(6). Ectopic expression of several PAX genes in NIH/3T3 cells induces cellular transformation and tumor formation in nude mice, suggesting that deregulated expression of PAX proteins could play a role in human tumorigenesis (6, 7). Furthermore,suppression of apoptosis by PAX proteins is crucial for their complex developmental role and could account for their tumorigenic potential(8, 9). In accordance with this, antisense inhibition of PAX genes results in growth arrest and apoptosis in tumor cell lines (10, 11, 12). Although evidence supports a role for PAX3 in protection of cells from apoptosis during development,the mechanism has not been determined (11).

The developmental abnormalities and alterations in gene expression observed in the splotch mouse model, which contains mutations in the PAX3 DNA-binding domains, has suggested several downstream target genes regulated by PAX3 (13, 14). Candidate targets include mi (15), myoD and myogenin (16), pax7 (11) and others. Similarly, transfection of the PAX3-FKHR fusion protein present in ARMS into heterologous cells has been shown to up-regulate the expression of pdgfr-αand c-met, the receptor for hepatocyte scatter factor(17). Whether any of these candidate genes play a role in ARMS tumorigenesis remains to be clarified.

It is generally hypothesized that the enhanced transcriptional activation potential of PAX3-FKHR is responsible for ARMS. We have previously engineered synthetic PAX3 repressors using the KRAB repression domain and demonstrated that expression of a PAX3-KRAB repressor in the ARMS Rh30 cell line could inhibit malignant growth(18). The KRAB domain functions as a potent DNA binding-dependent transcriptional repression module by recruiting the KAP-1 corepressor (19, 20). Other repression domains such as the SNAG domain from the GFI-1 proto-oncogene(21) and the WT-1 repression domain derived from the Wilms’ tumor gene (22) do not use the KAP-1 corepressor mechanism. The KRAB and the SNAG domains are well suited for the creation of engineered repressors due to their small size and strong repression potentials when fused to heterologous DNA-binding domains.

We were interested in developing conditional PAX3 repressors to examine the immediate consequences of repressing PAX3 target genes in ARMS cell clones. The use of conditional repressors avoids secondary changes in cells that might be selected by constitutive expression of transcriptional repressors. Several conditional eukaryotic expression systems based on either inducible transcription or conditional activity due to fusion to the HBD of steroid receptors have been developed(23). Of these, the HBD fusion confers rapid temporal regulation to the functionality of heterologous proteins. Furthermore,specific mutations in the HBDs make these receptors very selective to synthetic ligands without being influenced by endogenous hormones(24). HBD fusions with transcription factors including PAX5 (25, 26, 27) and enzymes such as STAT6 (28)have been successfully made to generate hormone-dependent conditional alleles.

In this study, we generated hormone-inducible, conditional alleles of a KRAB-PAX3 protein by fusing it to the HBD of the murine estrogen receptor ER^TM, which exhibits selectivity to 4-OHT (24). Hormone-dependent changes in biological properties such as growth in low-serum medium, apoptosis,anchorage-independent growth, and growth as tumors in SCID mice, were studied using ARMS Rh30 cell clones expressing the KRAB-PAX3-HBD repressors. The results of these experiments suggest that we have successfully used the inducible repressor strategy to down-regulate the set of PAX3 target genes that are activated by the PAX3-FKHR oncoprotein. Furthermore, we have used the conditional PAX3 repressors in ARMS cells to explore whether the cellular survival factor BCL-X_L might be a PAX3 target gene and be involved in ARMS tumorigenesis. These studies represent a first step in the identification of important oncogenic targets in ARMS by creation of a biological system well suited for analysis using differential gene expression array technologies.

The NIH/3T3 cell line was maintained in DMEM supplemented with 10%calf serum, 2 mm glutamine, 100 IU/ml penicillin, and 100μg/ml streptomycin at 37°C in 5% CO₂ under sterile conditions. The COS-1 cells were grown in Iscove’s modified Dulbecco’s medium containing 10% FBS and other components, as above. The parental Rh30 cell line and its clonal derivatives were maintained in RPMI 1640 containing 10% calf serum supplemented as above.

The previously described pcDNA3-PAX3-STOP plasmid, which expresses a hybrid mouse-human PAX3 protein, was used as a base to construct the KRAB-PAX3 wild-type and mutant fusion genes (4, 18). This plasmid was digested with HindIII and BamHI and ligated to the wild-type or the mutant(D₁₈V₁₉ changed to A₁₈A₁₉) KRAB repression domains (19). The KRAB domain-encoding fragments were generated by PCR amplification from the pM1-KOX-1 template(29) and were derived as HindIII and BamHI fragments encoding amino acids 1–90 of the KOX1 cDNA. The 5′ oligonucleotide primer incorporated a HindIII site and a Kozak consensus immediately before the KOX-1 initiator methionine (5′ primer, 5′-TTTTAAGCTTCCACCATGGATGCTAAGTCAC-3′). The 3′oligonucleotide primer incorporated a BamHI site after amino acid 90 of KOX-1 (3′ primer, 5′-TTTTGGATCCAGTCTCTGAATCAGGATG-3′). The resulting pcDNA3-KRAB-PAX3-STOP plasmid contains the KRAB domain,followed by a small linker encoding amino acids GSGVP, followed by amino acids 11–381 of the PAX3 DNA-binding domain. After amino acid 381, the PAX3-STOP protein is terminated by a vector-derived stop codon(18). The pcKRAB-PAX3-HBD plasmid was constructed by fusing the HBD, ER^TM of the murine estrogen receptor, to pcKRAB-PAX3-STOP. The ER^TM DNA was generated by PCR amplification from the pBS⁺ER^TM plasmid(24) template using a pair of oligonucleotide primers designed to incorporate flanking EcoRI sites (5′ primer,5′-GCATGAATTCTATGGGTGCTTCAGGAG-3′; 3′ T3 promoter primer,5′-AATTAACCCTCACTAAAGGG-3′). The PCR product was digested with EcoRI and ligated to the unique EcoRI site just 5′ of the vector-derived stop codon in the pcKRAB-PAX3-STOP plasmid to create an in-frame fusion. The pcSNAG-PAX3-HBD plasmid was constructed by fusing the HBD to the pcDNA3-SNAG-PAX3 plasmid, as described above. The fragment encoding the SNAG domain was generated by overlapping PCR amplification. The 5′ primer was designed to incorporate an EcoRI and a BamHI site, a Kozak consensus sequence, and the SNAG domain sequences (encoding amino acids 1–15; 5′primer, 5′-GAATTCGGATCC ACCATGCCACGTTCTTTCCTGGTTAAATCTAAAAAAGCGCACTCTTACC-3′). The3′primer contained the remaining portion of the SNAG domain (amino acids 16–20 in an antisense orientation), followed by BglII and SalI sites (3′-primer,5′-GTCGACAGATCTGGAGTAGTCCGGACCCGGAGAACGCGGCTGGTGGTAAGAGTGCGCTTTTTTAG-3′). These two oligonucleotides were annealed and amplified to yield a 105-bp fragment that was used as a template in the PCR reaction with a pair of flanking primers: (5′-GTCAGAATTCGGATCCACC-3′; and 3′ primer,5′-CCAAGTCGACAGATCTGGAG-3′). The resulting PCR product was digested with BamHI and BglII and cloned into the BamHI site in the pcDNA3-PAX3-STOP plasmid. The general-purpose vector pcDNA3-KRAB-MNP-HBD was constructed in two steps. First, the KRAB domain was amplified using the 5′ HindIII primer described above and a 3′ primer that incorporated a myc-epitope tag (EQKLISEEDL) and a nuclear localization signal of the E1A gene (KRPRP) immediately after amino acid 89 of the KRAB domain, followed by a BamHI site. Next, the ER^TM DNA was generated by PCR amplification using a 5′ primer designed to incorporate a polylinker composed of BamHI, EcoRI, EcoRV, and ClaI sites, and the 3′ T3 primer. This fragment was cleaved with BamHI and NotI, and then these two fragments were cloned into the HindIII and NotI sites of the pcDNA3 vector to generate the final construct. The nucleotide sequences of all PCR-derived constructs were confirmed by sequencing both strands. The previously described PAX reporter plasmid,CD19–2(A-ins)-TK-LUC (30), was modified by incorporation of a Zeocin^R cassette derived as a PvuII fragment from the pcDNA3.1Zeo plasmid (Invitrogen) to create the CD19–2(A-ins)-TKLUC-Zeo^Rplasmid.

The expression plasmids depicted in Fig. 1 A were transfected into COS-1 cells for 6 h with a mixture of DNA:lipofectAMINE in the ratio 1:6 in optiMEM, followed by growth for 48 h in DMEM containing 10% FBS. Transfected cells were metabolically labeled with ³⁵S-methionine,and the cell extracts prepared in radioimmunoprecipitation assay buffer were subjected to immunoprecipitation analysis with α-KRAB, α-PAX3,and α-HBD antibodies, as described previously (4, 18).

The DNA-binding potentials of KRAB-PAX3 and KRAB-PAX3-HBD proteins were assessed by EMSA performed using ³²P-labeled e5 DNA probe, as described previously (4, 31). The KRAB-PAX3 and KRAB-PAX3-HBD proteins used for EMSA were synthesized using the Promega TnT T7 transcription and translation system. In parallel reactions, ³⁵S-labeled proteins were prepared and analyzed by 10% SDS-PAGE and fluorography to confirm and normalize specific in vitro synthesis. For the corepressor supershift studies, COS-1 nuclear extract was used as a source of the KAP-1 corepressor (∼3 μg/μl, prepared as described previously; Refs.4 and 31). For the antibody supershift studies, α-KRAB or α-PAX3 IgG (2 μg/μl) was included in the EMSA-binding reactions.

Total RNA was isolated from Rh30-KPHBD-cl22 cells using the TRIzol reagent (Life Technologies, Inc., Rockville, MD) and electrophoresed on 1% formaldehyde-agarose gels. Prior to RNA isolation, Rh30-KPHBD-cl22 cells were either un-induced (treated with 0.1% ethanol as a solvent control) or induced with 500 nm 4-OHT. The gel was stained with ethidium bromide to assess the integrity and equal loading of the samples and then transferred to a nylon membrane (Hybond). The membrane was prehybridized, hybridized with radiolabeled PAX3-KRAB probe, and washed to a final stringency of 0.2× SSC, 0.2% SDS, at 65°C, before autoradiography.

Subcellular localization of KRAB-PAX3 and KRAB-PAX3-HBD proteins in Rh30, Rh30-pcDNA-cl5, and Rh30-KPHBD-cl22 cells grown on cover glasses was conducted using previously described procedures for immunofluorescence (31). After fixation with 1%paraformaldehyde and permeabilization with 0.2% Triton X-100 (Sigma Chemical Co.), the antigens were localized using α-PAX3 or α-HBD primary rabbit antibodies, followed by detection with a secondary biotinylated α-rabbit IgG and avidin-FITC (Vector Laboratories,Inc.). The nuclei were counterstained for DNA with 0.5 μg/ml Hoechst 33258 (Sigma Chemical Co.) and the cells were visualized using a Leica confocal laser-scanning microscope. The HC-20 antibody to the HBD of the murine estrogen receptor was obtained from Santa Cruz Biotechnology Inc.

The transcription assays were performed on NIH/3T3 cells that were transiently transfected with a LipofectAMINE mixture containing 1 or 2.5 μg of the expression plasmids (pcDNA3, pcKRAB-PAX3, and pcKRAB-PAX3-HBD), 0.5 μg of CD19–2(A-ins)-TK-LUC, and 0.25 μg of CMV-β-d-galactosidase plasmids, as described previously(18). Transfected cells were treated with 0.1% ethanol as a solvent control for un-induced dishes or were induced with 500 nm 4-OHT (Research Biochemicals International, Natick, MA). After 24 h, the cells were washed twice with Tris-buffered saline,and the cell extracts were prepared in reporter lysis buffer and assayed for luciferase and β-galactosidase activities as described(31).

Rh30 cell transfectants containing the conditional PAX3 repressor plasmids depicted in Fig. 1 A or the pcDNA3 vector were isolated as individual colonies using cloning rings and were expanded into cell lines after selection for stable resistance to 500 μg/ml G418 (Mediatech, Inc., Herndon, VA). Twenty-four independent cell lines were tested for expression of the PAX3-HBD repressor proteins by immunoprecipitation with α-PAX3 IgG. Dual-stable inducible PAX3 repressor/PAX3 reporter cell clones were generated in the Rh30-KPHBD-cl22 cell line after transfection with the CD19–2(A-ins)-TK-LUC-Zeo^R plasmid and selection with 500 μg/ml G418 and 100 μg/ml Zeocin. The PAX3 repressor/PAX3 reporter cell lines, designated as HBDLUC clones, were screened for repression of luciferase after induction with 4-OHT. The luciferase activities of the HBDLUC clones were normalized to a protein content of 1.0 A595 unit in the BioRad protein assay.

Gene expression profiles of un-induced and 4-OHT-induced Rh30-KPHBD-cl22 cells were analyzed using DD-PCR. Total RNA was isolated using the TRIzol reagent and poly(A)⁺mRNA was purified using the oligo-(dT)₂₅Dynabeads (Dynal, Inc., Lake Success, NY). First-strand cDNA was synthesized using the Life Technologies, Inc. cDNA synthesis system and quantitated by spectrophotometry. Differential subtraction display PCR was conducted as described (32), and the samples were electrophoresed on a sequencing gel and autoradiographed.

The growth assays of Rh30, Rh30-KPHBD-cl22, Rh30-SPHBD-cl8, and Rh30-K(DV)PHBD-cl24 cell lines in low-serum (0.1% FBS) medium was conducted in 24-well tissue culture plates and was repeated twice with at least 10 replicates. The anchorage-independent growth of parental Rh30 and Rh30-KPHBD-cl22 cells was evaluated using poly-HEMA-coated plates. In both assays, the proportion of viable cells at each time point was determined by MTT assay, as described previously(33).

Apoptosis assays were performed using the ApoAlert DNA Fragmentation and the ApoAlert Annexin V Assay kits according to the manufacturer’s instructions (Clontech Laboratories, Inc.). Assays were conducted on SCID mouse tumor sections or on Rh30-KPHBD-cl22 cells that were grown on coverslips in low-serum medium under either un-induced (0.1%ethanol) or 4-OHT-induced conditions (500 nm, 48 h). The polyclonal rabbit antibody for immunoblot detection of BCL-X_L (bcl-x, Ab-1) and the antibody for detection of human α-tubulin as a loading control were obtained from Oncogene Research Products (Cambridge, MA).

For RT-PCR analysis, RNA was isolated from the un-induced or the 4-OHT-induced Rh30-KPHBD-cl22 and Rh30-SPHBD-cl8 cells grown in low-serum medium using the TRIzol reagent. The reverse transcription reactions were performed on 5 μg of total RNA using oligo-dT primers with the Ready-To-Go You-Prime First-Strand Synthesis Beads (Pharmacia Biotech). The PCR reactions were carried out with 2.5 μl of reverse transcription reactions as templates. A pair of primers specific for the human BCL-X_L transcript (5′ primer,5′-CAGCAGCAGTTTGGATGC-3′; 3′-primer, 5′-CCACAGTCATGC CCG TC-3′) was used to amplify the 448-bp product. A specific primer pair (5′-primer,5′-TCAGCGCAGGGGCGCCCGGTTCTT T-3′; and 3′-primer,5′-ATCGACAAGACCGGCTTCCATCCGA-3′) was used to amplify the 345-bp product from the Neo^R gene. A pair of primers specific for the HBD of the murine estrogen receptor (5′-primer,5′-GCGACGGGCCCATGGGTGCTTCAG G-3′; and 3′-primer,5′GGTGGGCCCCTGATATCACAAGTCCTCTTCAGAAATGAGCTTTTGCTCGATCGTGTTGGGGAAGCC-3′)was used to amplify a 1010-bp product from the SCID mouse tumor samples that were derived from Rh30-KPHBD-cl22 cells.

The tumorigenic potentials of Rh30-pcDNA-cl5 and Rh30-KPHBD-cl22 cell lines were evaluated after s.c. injection into female CB17-SCID mice, 6 weeks of age. Evidence of tumor growth became apparent after 10 days,at which time the mice were divided into two groups of five mice each(un-induced and 4-OHT induced). The mice of the 4-OHT-induced group were implanted with 35-mg timed-release pellets specified to maintain a 200-nm circulating concentration of 4-OHT for 21 days(Innovative Research of America, Sarasota, FL). Improved implant success was ensured by application of DERMABOND topical skin adhesive(Ethicon, Inc., Somerville, NJ) over the wounds. Alternate day measurements of tumor volumes were made using a tumorimeter (Cancer Technologies, Inc., Tucson, AZ). After 3 weeks, the mice were sacrificed and the wet weights of the tumors were recorded. A portion of each tumor was fixed in formalin for H&E staining and histopathological evaluation.

We have previously demonstrated that we could engineer a PAX3 transcriptional repressor by fusing the KRAB domain to the PAX3 DNA-binding domains. We found that constitutive expression of a PAX3-KRAB protein could inhibit the malignant growth of ARMS cells(18). In the present study, a KRAB-PAX3 plasmid was converted to a conditional repressor, KRAB-PAX3-HBD, by fusion of the ER^TM domain in-frame to the COOH terminus (Fig. 1,A). The ER^TM domain is a well-established, tamoxifen-selective, mutant version of the HBD of the murine estrogen receptor that confers hormone-dependent functionality to heterologous fusion proteins (24). The KRAB(DV)-PAX3-HBD protein features a mutant KRAB repression domain that lacks transcriptional repression potential and is a useful control to confirm that the elicited biological responses depend on a functional repression module (19). Furthermore, by stable integration of the 6× CD19–2(A-ins)-TK-LUC luciferase reporter plasmid (Fig. 1,B), we have created inducible PAX3 repressor/reporter(HBDLUC) cell lines that permit convenient monitoring of the repression function in a chromatin-mediated state. The modular nature of the engineered repressor is shown by the use of another conditional repressor plasmid, SNAG-PAX3-HBD, which was created using the SNAG repression module of the GFI-1 gene(21).⁴Clearly, the use of engineered repressors is not limited to the set of target genes selected by the PAX3 DNA-binding domains as in this study. The general purpose KRAB-MNP-HBD vector (Fig. 1 C) was designed for use with other DNA-binding domains of interest that could be inserted into the poly-linker between the KRAB repression domain and the ER^TM domain. This plasmid features incorporation of a nuclear localization signal and a c-mycepitope tag to aid in detection. The resulting conditional chimeric repressors can be applied for repression of the set of endogenous target genes of choice determined by the inserted DNA binding motif.

Immunoprecipitation analysis of transfected COS-1 cell extracts using both α-KRAB and α-PAX3 IgGs (Fig. 2,A) demonstrates that the pcKRAB-PAX3 (55 kDa) and pcKRAB-PAX3-HBD (88 kDa) proteins are expressed in vivo at the size predicted by their cDNAs and are recognized by the appropriate antibodies. The EMSA analysis (Fig. 2,B) indicated the presence of clear binary complexes of both KRAB-PAX3:e5DNA and KRAB-PAX3-HBD:e5DNA, which confirmed that both KRAB-PAX3 and KRAB-PAX3-HBD proteins exhibited the anticipated DNA-binding properties. We have previously shown, by competition with unlabeled homologous oligonucleotide, that binding of PAX3 protein to the e5 binding site probe is specific (4). Furthermore, these binary complexes were confirmed to contain the predicted proteins because they were efficiently super-shifted when α-KRAB or α-PAX3 IgG was included during the DNA-binding reactions (data not shown). It was also evident that the KRAB domain present in the binary complex was amenable for protein-protein interactions because ternary complexes of KRAB-PAX3:KAP-1:e5DNA or KRAB-PAX3-HBD:KAP-1:e5DNA were observed when COS-1 nuclear extract was included in the gel-shift reaction as a source of the KAP-1 corepressor (20). We have also found that the SNAG-PAX3-HBD protein efficiently binds the e5 DNA probe but does not form a ternary complex with KAP-1 (data not shown). As would be expected with the ER^TM posttranslational regulation system, we have shown that similar levels of KRAB-PAX3-HBD transcript are expressed in Rh30-KPHBD-cl22 cells grown under both un-induced and 4-OHT-induced conditions (Fig. 2,C). Immunofluorescence microscopy indicated that in the absence of hormone the KRAB-PAX3-HBD protein was predominantly cytoplasmic (Fig. 10,A, top), but after induction with 4-OHT it was located in the nucleus (Fig. 10 A, bottom).

Transcriptional repression assays that were carried out in NIH/3T3 cells transiently transfected with the KRAB-PAX3 plasmid showed constitutive repression that depended on the dose of the input expression plasmid but not on the presence of 4-OHT. In contrast, cells transfected with 1 μg of the KRAB-PAX3-HBD plasmid exhibited a significant dependence on the presence of 4-OHT to repress the luciferase reporter and typically manifested a 10-fold repression (Fig. 3). A hormone-independent repression was also observed at a high input dose (2.5 μg) of KRAB-PAX3-HBD plasmid. As expected, the vector-transfected cells showed no repression (Fig. 3). Overall, our characterization of the KRAB-PAX3-HBD repressor indicated proper expression, DNA-binding activity, and a significant degree of hormone-dependent transcriptional repression activity in NIH/3T3 cells. NIH/3T3 cells do not express PAX3, hence, lack any endogenous PAX3-dependent transcriptional activation function (data not shown;Ref. 34).

Our primary interest, however, was to evaluate the conditional PAX3 repressor in Rh30 ARMS cell lines that express the endogenous PAX3-FKHR oncogenic transcriptional activator. Therefore, we generated stable Rh30 cell clones, which expressed the conditional PAX3 repressors and then evaluated their repression potentials after transient transfection of the CD19–2(A-ins)-TK-LUC reporter plasmid. In contrast to the 10-fold repression observed in transiently transfected NIH/3T3 cells(Fig. 3), in Rh30 cells the repression achieved ranged from 2–4-fold,consistent with the presence of a competing endogenous activation due to PAX3-FKHR. Hormone-dependent repression of the reporter plasmid was observed in all of the Rh30-KPHBD clones (e.g., −cl11,cl13, cl19, and cl22; Fig. 4,B) that correlated well with the expression levels of the KRAB-PAX3-HBD fusion protein (Fig. 4,A). The observed conditional transcriptional repression appeared to be dominant over transcriptional activation elicited by the endogenous PAX3-FKHR oncogene. Similarly, all of the SNAG-PAX3-HBD-expressing Rh30 cell clones (e.g., −cl4, cl7,cl8, and cl24; Fig. 4,F) showed a good correlation between hormone-dependent repression of luciferase activities and expression levels of the SNAG-PAX3-HBD protein (Fig. 4,E). As expected,the Rh30 cell clones (e.g., −cl3, cl13, and cl24; Fig. 4,C) that expressed the mutant KRAB(DV)-PAX3-HBD repressor protein (Fig. 4,C) showed no repression (Fig. 4 D). Thus, repression was found to completely depend on the presence of a functional repression domain (KRAB or SNAG) but was not restricted to a particular type of repression mechanism.

Our intended goal was to apply the PAX3 conditional repressor strategy to specifically down-regulate endogenous target genes that are activated by the PAX3-FKHR oncogene. Hence, we produced stable cell clones containing both the KRAB-PAX3-HBD plasmid(Neo^R) and the CD19–2(A-ins)-TK-LUC reporter plasmid (Zeo^R), designated as HBDLUC cell lines. This allowed us to correlate repression of an integrated artificial reporter gene target with reversion of ARMS malignant growth characteristics and changes in expression of endogenous PAX3target genes. Several independent clonal cell lines were tested for hormone-dependent repression of the chromatin-integrated luciferase gene, and the results obtained for a selected number of clones are depicted in Fig. 5,A, which shows that the normalized luciferase activities varied from 10³–10⁶luciferase units. The fold repression ranged from >1.5–5 for different clonal cell lines (Fig. 5,B). The variation observed between clones most likely reflects differences in reporter copy number and integration site. The range of repression potentials obtained with the chromatin-integrated reporter clones are consistent with those observed in transient assays in Rh30 cells (Fig. 4), further supporting the contention that the PAX3 repressors could dominantly repress endogenous genes despite the presence of the PAX3-FKHR protein.

Our studies have indicated that repression of PAX3-FKHR target genes is a strategy for reversion of malignant growth of ARMS cells. Thus, we have established a relevant system to model the conditional repression of endogenous PAX3-FKHR target genes. To demonstrate that hormone-dependent repression of endogenous genes was occurring in this ARMS cell system, DD-PCR analysis was performed using RNA made from un-induced and 4-OHT-induced Rh30-KPHBD-cl22 cells. When different combinations of anchor and arbitrary primers were used, the DD-PCR products displayed significant expression pattern differences (data not shown). Isolation and characterization of these differentially expressed mRNA species is currently under investigation, and candidate genes are being evaluated for functional relevance to the malignant phenotype of ARMS.

To further substantiate the inducible PAX3 repressor Rh30 cell lines as pertinent model systems for the study of PAX-3-FKHR target genes in ARMS, we used the Rh30-KPHBD-cl22 cell line to examine whether KRAB-PAX3-HBD would function as a conditional repressor of ARMS malignant growth. Growth properties of Rh30 and Rh30-KPHBD-cl22 cells were studied in the presence or absence of 4-OHT under full-serum(10%) and reduced-serum (0.1%) conditions. The growth in full serum was not significantly changed relative to growth of a parallel un-induced culture on activation of KRAB-PAX3-HBD protein by 4-OHT(data not shown). However, under low-serum conditions, a significant reduction in the number of viable Rh30-KPHBD-cl22 cells over a 10-day period was observed when the KRAB-PAX3-HBD repressor was activated by 4-OHT and growth was evaluated by MTT assay (Fig. 6,B). Rh30-SPHBD-cl8 cells expressing the SNAG-PAX3-HBD protein manifested similar hormone-dependent inhibition of cell growth (Fig. 6,C). As expected, Rh30-K(DV)PHBD-cl24 cells did not show any growth retardation under either conditions (Fig. 6,D),similar to parental Rh30 cells (Fig. 6,A). Growth of un-induced Rh30-KPHBD-cl22 cells (Fig. 6,B) was also similar to that of Rh30 cell line (Fig. 6,A). These results clearly indicate that conversion of the inactive repressor protein to an active form by 4-OHT is responsible for this growth inhibition in Rh30-KPHBD-cl22 and Rh30-SPHBD-cl8 cells. The malignant phenotype of the Rh30-KPHBD-cl22 cells was further examined in poly-HEMA-coated tissue culture plates to evaluate anchorage-independent growth potential. As expected, the parental Rh30 cell line exhibited abundant growth under the conditions of the poly-HEMA assay, and the growth was equivalent under un-induced and 4-OHT-induced conditions (Fig. 6,E). In marked contrast, the Rh30-KPHBD-cl22 cell line showed a dramatic suppression of growth potential only on 4-OHT treatment (Fig. 6 F).

Our data indicate that activation of the KRAB-PAX3-HBD repressor renders the Rh30-KPHBD-cl22 cells very sensitive to inhibition of growth in low-serum medium as evidenced by a loss of cell viability by day 8 (Fig. 6,B). The PAX3 repressor also inhibited growth under anchorage-independent conditions as seen above, however, there was no evidence of drastic cell loss. We were interested to know whether the difference in growth kinetics observed between the two different assays could be due to apoptosis. Hence, we carried out two independent apoptosis assays (TUNEL and Annexin V) using Rh30-KPHBD-cl22 cells that were grown in low serum under un-induced or 4-OHT-induced conditions. On 4-OHT induction, the TUNEL assays showed a significant increase in the proportion of cells stained with streptavidin-FITC (Fig. 10 B). This enhanced staining reflects incorporation of biotin-16-dUTP at DNA strand breaks and is a sensitive indication of apoptotic cell death. The level of nuclear staining by propidium iodide was the same in the presence or absence of 4-OHT induction, ruling out nonspecific cytotoxicity (data not shown). Enhanced immunodetection of Annexin V was also observed after 4-OHT induction (data not shown), supporting the correlation between KRAB-PAX3-HBD activity and apoptosis in low-serum culture.

These findings suggested that the KRAB-PAX3-HBD protein might have repressed a PAX3 target gene linked to protection from apoptosis,consistent with the role described for PAX proteins during development(10, 11). It has previously been reported that PAX8 can promote cell survival by activating antiapoptotic factors of the BCL family (8). We evaluated expression of the BCL-X_L survival factor in Rh30-KPHBD-cl22 cells that were grown in low serum under un-induced or 4-OHT-induced conditions. Immunoblot analysis of whole cell lysates indicated that the BCL-X_L protein levels from 4-OHT-induced cells were decreased compared with un-induced cells (Fig. 7,A). RT-PCR analysis using BCL-X_L-specific primers indicated a reduced amplification of the BCL-X_L product from reactions from 4-OHT-induced samples compared with the un-induced samples. These data indicate that activation of KRAB-PAX3-HBD protein results in repression of BCL-X_L transcription(Fig. 7 B).

To test whether the KRAB-PAX3-HBD protein is a direct repressor of BCL-X_L, Rh30-KPHBD-cl22 cells were induced with 4-OHT for 10 h in the presence of cycloheximide. The RT-PCR analyses (Fig. 7 C) indicated that amplification of the BCL-X_L product continued to be diminished compared with the un-induced controls. However, the BCL-X_L product generated from the cycloheximide-treated sample was not repressed as fully as the 4-OHT-induced, noncycloheximide-treated sample. This diminution may be attributed to a reduced level of the KRAB-PAX3-HBD protein after the cycloheximide treatment (data not shown). Overall, these results suggest that repression of the BCL-X_L product does not require de novo protein synthesis and BCL-X_L is a probable candidate for a direct target gene of PAX3-KRAB-HBD.

Our previous studies had demonstrated that PAX3-KRAB expression could suppress Rh30 cell tumorigenesis in SCID mice (18). Thus,we were interested in whether conditional suppression of tumorigenesis in vivo could be demonstrated using the newly established conditional PAX3 repressor cell line Rh30-KPHBD-cl22. We generated tumors by injecting SCID mice with Rh30-pcDNA-cl5 or Rh30-KPHBD-cl22 cells. After a 10-day period to establish tumor growth, half of the mice having comparably sized tumors were implanted with slow-release pellets containing the 4-OHT inducer. Subsequently, tumor measurements were made at regular intervals for 3 weeks, and these results over time are shown in Fig. 8,A for Rh30-pcDNA-cl5 mice and in Fig. 8,B for Rh30-KPHBD-cl22 mice. Data for the measured tumor volumes are presented in Fig. 8,C. After sacrifice, tumors were isolated by dissection, and their wet weights are plotted in Fig. 8,D. These results agree with the tumor volume estimates and clearly indicate that in the presence of 4-OHT only the Rh30-KPHBD-cl22 mice showed a reduction in tumor growth, whereas the Rh30-pcDNA-cl5 mice did not. The mice bearing the Rh30-pcDNA-cl5 and the Rh30-KPHBD-cl22 tumors were photographed before sacrifice (Fig. 9,A). RT-PCR analysis of tumor RNA using the HBD 5′ and 3′primers indicated that only Rh30-KPHBD-cl22 (-/+ 4-OHT)-injected mice expressed the KRAB-PAX3-HBD transcript (Fig. 9,B). Furthermore, the tumors from the un-induced or 4-OHT-induced pcDNA-cl5 mice and the un-induced KPHBD-cl22 mice contained many blood vessels in the H&E-stained tumor tissue sections (Fig. 10,C). On the contrary, in the 4-OHT-induced KPHBD-cl22 mice significantly fewer blood vessels were evident. It is presently unknown whether the reduced angiogenesis is directly related to the KRAB-PAX3-HBD protein or a secondary consequence of the reduced tumor burden. TUNEL assays performed on the tumor sections from the negative control mice (un-induced or 4-OHT-induced pcDNA-cl5 mice and the un-induced KPHBD-cl22 mice) showed a similar low level of staining,indicating very little apoptosis in vivo (Fig. 10 D). However, an enhanced streptavidin-FITC staining was observed in the tumor sections from the 4-OHT-induced KPHBD-cl22 mice,indicative of KRAB-PAX3-HBD-induced apoptotic cell death in vivo.

In this study, we have generated ARMS cell lines with conditional PAX3 repressors to establish a system to understand the biological role and identify endogenous target genes of the PAX3-FKHR oncoprotein. The conditional PAX3 repressors were generated by fusing the KRAB or SNAG repression domains and PAX3 DNA-binding motifs to the tamoxifen-selective ER^TM HBD of the estrogen receptor. We believe that application of our conditional repressors to the ARMS model system is especially relevant for evaluating the cellular pathways aberrantly regulated by the PAX3-FKHR protein. One important advantage of using the ER^TM system is that rapid activation of the transcription factor-ER fusion protein can be achieved by a simple ligand-mediated on/off mechanism (24, 35). Thus, the immediate consequences of the KRAB-PAX3-HBD repressor on biological properties contributing to malignant growth and tumorigenesis can be studied without selection for secondary mechanisms.

Our confidence in the conditional repressors applied in this study is supported by the following principal results that support their effective dominant negative function. First, the KRAB-PAX3-HBD and SNAG-PAX3-HBD proteins were fully competent in binding to the e5 target DNA in EMSA analysis. The super-shift experiments showed that the KRAB-PAX3-HBD protein was also fully capable of binding the KAP-1 corepressor, which is a prerequisite for KRAB-mediated transcriptional repression (20). Conditional repressors using the SNAG domain do not rely on KAP-1 association, thus, may have an extended applicability in biological systems in which KAP-1 is not present, such as in insect cells, or in certain differentiated cell types(31). Second, the ability of the KRAB-PAX3-HBD and SNAG-PAX3-HBD conditional repressors to induce transcriptional repression was tightly regulated by the ligand, 4-OHT. Moreover, the degree of transcriptional repression achieved by the conditional KRAB-PAX3-HBD repressor was as potent as the repression observed with the constitutive KRAB-PAX3 repressor, implying that fusion of the ER HBD did not interfere with the expected function. Third, in Rh30 stable clones, the KRAB-PAX3-HBD and the SNAG-PAX3-HBD protein expression levels were higher than that of the endogenous PAX3-FKHR protein;hence, the repressors would be expected to have a competitive advantage for PAX3-FKHR target genes. Fourth, the conditional PAX3 repressor proteins exhibited hormone-dependent nuclear localization, which may account for the hormone-dependent biological effects in Rh30 cells. Fifth, the KRAB-PAX3-HBD and SNAG-PAX3-HBD proteins exhibited potent repression of either transiently transfected or integrated reporter genes in Rh30 ARMS cells. The conditional repressors were as effective in repression of integrated reporter plasmids as transiently transfected reporter plasmid. Furthermore, the repression of endogenous genes by the conditional repressors was confirmed by the DD-PCR analysis. In all cases, the conditional repression was dominant over any activation mediated by the endogenous PAX3-FKHR. In addition, there was a striking correlation between the expression levels of the PAX3 repressors and the degree of transcriptional repression. Sixth, the conditional PAX3 repressors exhibited hormone-dependent suppression of characteristic malignant properties such as: (a) growth in low serum; (b) anchorage-independent growth; and(c) tumorigenicity in SCID mice. Furthermore, both the KRAB-based as well as the SNAG-based conditional PAX3 repressors exhibited similar inhibitory properties, demonstrating that the growth inhibition did not depend on a particular type of repression mechanism. Finally, the temporal control afforded by the conditional PAX3 repressor allowed observation of immediate changes in growth under low-serum conditions that led us to demonstrate that the KRAB-PAX3-HBD protein induces apoptosis.

The hormone-dependent apoptosis exhibited in the conditional PAX3 repressor Rh30 cell lines under low-serum conditions is entirely consistent with previous studies in which an antisense approach was used to down-regulate PAX3-FKHR in Rh30 ARMS cells (10). Furthermore, it has been clearly demonstrated that down-regulation of PAX3 function, either by mutation in splotch mice(14), or by reduced expression in diabetic mice(36), leads to extensive apoptosis. It is well established that many of the PAX family proteins function during organogenesis to protect cells from apoptosis (9, 11, 14). Previous studies have linked the antiapoptotic function of PAX8 to activation of the BCL family protein, BCL-2 (8). Our studies have demonstrated that the KRAB-PAX3-HBD protein can down-regulate expression of the BCL-X_L cellular survival factor and induce apoptosis in Rh30 cell lines in vitro and in SCID mice in vivo. These observations suggest that PAX3-FKHR may contribute to oncogenesis in ARMS by activating the antiapoptotic BCL-X_L survival factor. One advantage of the ER system for studying transcription factors like KRAB-PAX3-HBD is that activation of the ER-fusion protein can occur in the absence of de novo protein synthesis. Thus, in the presence of protein synthesis inhibitors such as cycloheximide, a distinction can be made between directly regulated target genes and those secondary targets whose regulation depends on transcription and translation of a primary target gene. Our data supports the hypothesis that BCL-X_L is a direct PAX3 target gene,because the repression by KRAB-PAX3-HBD was largely insensitive to inhibition by cycloheximide. Recent studies have confirmed that PAX3 and PAX3-FKHR can directly regulate the BCL-X_Lvia direct binding to the promoter (37).

It is likely that regulation of apoptosis by PAX3 may involve several other mechanisms in addition to BCL-X_L. Recent expression profiling has shown that PAX3-FKHR can activate expression of the SLUG protein (34). Other studies have shown that SLUG plays an antiapoptotic role in E2A-HLF-induced tumorigenesis(38). In addition, PAX3 is known to regulate c-MET, the receptor for the hepatocyte growth factor/scatter factor, which has also been implicated in apoptosis (39, 40). It is interesting that apoptosis was only observed when the KRAB-PAX3-HBD protein was activated under low-serum conditions, implying that the Rh30 cells may only depend on PAX3-FKHR for protection against apoptosis when deprived of certain factors present in serum. Substantial evidence supports a role for growth factor signaling in ARMS. PAX3-FKHR has been shown to activate expression of insulin-like growth factor II, which has a well-established role in ARMS malignant growth (34, 41). Other studies have defined a role for the wild-type FKHR protein in linking insulin signaling to a cellular proapoptotic response (41, 42). It is not known whether PAX3-FKHR exerts any dominant-negative effect on the wild-type FKHR protein derived from the nontranslocated allele in ARMS. Protection from apoptosis by PAX3-FKHR is likely to be a key oncogenic mechanism in ARMS that will require further identification and validation of the essential target genes.

Our demonstration that KRAB-PAX3-HBD Rh30 cell tumors show pronounced hormone-dependent inhibition of growth confirms our previous findings and emphasizes that in ARMS the tumorigenic potential is dependent on PAX3-FKHR. In SCID mice experiments, use of the timed-release pellets allowed maintenance of the blood levels of 4-OHT and activation of the conditional repressor throughout the course of the experiment. The tumors that grew from control cells without the repressor or from Rh30-KPHBD-cl22 cells without the 4-OHT inducer were heavily vascularized. It is interesting that only the Rh30-KPHBD-cl22 tumors derived from the 4-OHT-treated mice showed very little evidence of vascularization in the H&E-stained tumor sections. We are currently investigating whether the KRAB-PAX3-HBD protein may have repressed angiogenic factors.

We believe that our conditional PAX3 repressor strategies have targeted a major oncogenic mechanism in ARMS, protection from apoptosis by PAX3-FKHR. Characterization of repressed target genes that comprise the PAX3-FKHR “oncogenic transcriptome” in ARMS identified by differential display RT-PCR analysis is our current focus.

We are grateful to Dr. Satish Parimoo (Skin Biology, Johnson &Johnson, Skillman, NJ) for providing reagents and guidance with DD-PCR protocols. We thank Dr. Beat Schafer for the 6xCD19–2 (A-ins)-TK-LUC reporter plasmid. We thank Trevor Littlewood (ICRF, London) for the HBD-ER^™ plasmid. Richelle Takemoto and Rachel Beurmann of Dr. Meenhard Herlyn’s laboratory at the Wistar Institute provided assistance with tumorigenicity studies in SCID mice. We are grateful to members of Dr. Frank Rauscher’s laboratory for helpful suggestions and comments.